Gravitational Wave-Signal Recognition Model Based on Fourier Transform and Convolutional Neural Network

Area of Science:

• Astrophysics
• Signal Processing
• Machine Learning

Background:

• Gravitational wave detection is a significant advancement in astrophysics.
• Traditional signal recognition methods are becoming insufficient with technological progress.
• Novel approaches are required for accurate gravitational wave signal identification.

Purpose of the Study:

• To develop an advanced gravitational wave signal recognition model.
• To leverage Fourier transformation and convolutional neural networks (CNNs) for enhanced feature description.
• To improve the accuracy of identifying real gravitational wave signals automatically.

Main Methods:

• Gravitational wave time-domain signals were converted into 2D frequency-domain signal graphs.
• A CNN model was employed for feature recognition on these frequency-domain graphs.
• The impact of training data size and image filtering was assessed; Resnet101 was used for comparison.

Main Results:

• Frequency-domain signal graphs offer superior feature descriptions compared to time-domain signals.
• The developed CNN model achieved higher recognition accuracy than the Resnet101 comparative model.
• The new method demonstrated approximately 4% higher average recognition accuracy.

Conclusions:

• CNNs are highly effective for gravitational wave signal recognition, building on their image recognition success.
• The proposed model provides a more appropriate and accurate method for automatic signal identification.
• This research paves the way for more reliable detection and analysis of gravitational wave events.